SK Hynix-Intel Talks Could Reuse Ohio’s Delayed $28B Fab
Intel slowed Ohio because demand did not justify the original construction pace. Memory scarcity now gives the site another use, but the economics depend on whether SK Hynix leases space, shares capital in a joint venture or buys manufacturing services.
SK Hynix is in exploratory talks with Intel about producing memory chips in the United States, Reuters reported, with possibilities including leasing part of Intel's Ohio site or forming a joint venture involving Intel and cloud customers. SK Hynix says nothing has been decided.
The asset already carries a large capital history. Intel committed more than $28 billion to two Ohio fabs, then slowed construction as part of a broader push to match spending with visible demand. Its current schedule puts the first fab into operation in 2030-31 and the second in 2032. In its latest filings, Intel still treats Ohio as a long-term funding requirement rather than near-term productive capacity.
That is why transaction structure matters more than the headline partnership. A lease could offset site carrying costs while keeping manufacturing risk with SK Hynix. A joint venture could bring in outside equity and split construction risk. A manufacturing-services contract could create operating revenue for Intel, but would also expose it more directly to process execution, yield and customer-volume commitments. The same Ohio square footage can produce very different returns.
Memory scarcity makes those options more valuable. Seoul Economic Daily reported a 64GB DDR5 server RDIMM at about $1,500 on September 15, up from $272 a year earlier, with some spot offers as high as $3,100. TrendForce had already forecast third-quarter server-DRAM contract prices rising 13% to 18% as suppliers prioritize server memory and HBM for AI workloads.
SK Hynix faces a different trade. US production would improve geographic resilience and put capacity closer to major cloud customers, but it is unlikely to be attractive at any cost. Subsidies, partner capital, customer commitments and the exact manufacturing arrangement will determine whether localization earns an acceptable return after higher US construction and operating costs.
Intel has already spent years creating the option represented by Ohio. Scarcity gives that option a buyer. A signed lease would mainly monetize dormant infrastructure; a JV would relieve capital intensity; a true manufacturing agreement would say more about Intel's ability to turn its US footprint into third-party production revenue. Until the structure is disclosed, those outcomes should not be collapsed into one bullish foundry narrative.